Skip to content

installing Rgraphviz

3 messages · Jarrett Byrnes, Patrick Aboyoun

#
I know variants of this have been answered a few times, but thus far  
I've had no luck:

I'm attempting to install Rgraphviz from bioconductor.  I downloaded  
the stablemac version from http://www.graphviz.org/Download_macos.php  
and used biocite to install Rgraphviz after installing bioconductor.   
However, when I attempt to load Rgraphviz, I get the following error.

Error in dyn.load(file, DLLpath = DLLpath, ...) :
   unable to load shared library '/Users/jbyr/Library/R/2.9/library/ 
Rgraphviz/libs/ppc/Rgraphviz.so':
   dlopen(/Users/jbyr/Library/R/2.9/library/Rgraphviz/libs/ppc/ 
Rgraphviz.so, 6): Library not loaded: /usr/local/lib/libgvc.3.dylib
   Referenced from: /Users/jbyr/Library/R/2.9/library/Rgraphviz/libs/ 
ppc/Rgraphviz.so
   Reason: image not found
Error: package/namespace load failed for 'Rgraphviz'


I went to look for libgvc.3.dylib, and there in /usr/lib found libgvc. 
5.dylib .

Is the answer here to install an older version of graphviz? If so  
which one?  Or can Rgraphviz be made to work with my existing  
installation of graphviz?

-Jarrett
#
Jarrett,
This is a Bioconductor related question and really should be addressed 
on the Bioconductor mailing list. The sort answer is if you are running 
on a Mac OS X 10.4 (Tiger) box, you will need to build both graphviz and 
Rgraphviz from source in order to use the package. If you are running on a
Mac OS X 10.5 (Tiger) box, continue reading.

The issue is that graphviz.org distributes their latest version of 
Graphviz for Mac OS X 10.5 (Leopard) only and we use that when we build 
our mac.binary.leopard version of Rgraphviz. I am guessing that you are 
using the cross Mac OS X pre-compiled version of R, which uses the 
mac.binary build of the package from our Mac OS X 10.4 (Tiger) build 
machine that has an old (unmaintained) version of graphviz on it. I 
recommend that all Bioconductor users running on leopard to use a 
leopard-only build of R available at http://r.research.att.com/, since 
that is what we use on our build machine. These leopard only R binaries 
download, via biocLite, mac.binary.leopard versions of the packages that 
are produced by our Leopard machine. These leopard only binaries have 
the added benefit of 64-bit Intel support so end-users don't have to 
build their own packages once they leave the 32-bit world.


Patrick
Jarrett Byrnes wrote:
#
Typo.

Change Mac OS X 10.5 (Tiger)

to

Mac OS X 10.5 (Leopard)

below.
Patrick Aboyoun wrote: